The key is that the soliders are set to a strategem, not a specific x-y-z position, and have to try to keep in that cadence, so to speak. Yoon Ha Lee's writing is incredibly strange and open to interpertation here, so if this sounds strange, it's somewhat meant to be (Calendrical Rot, corpselight, amputation guns, etc.). Reading the early chapters is better than me trying to explain it.
That said, trying to manage battle is fruitless and attempting to impose order into something that intentionally random and chaotic is not going to go well. Col. Boyd's OODA loop[1] is pertienet here. If your enemy knows that they have been looped, then putting random noise into the system can alievate that and get them back into the war.
It's a great little sequence, worth a read if you're into that sort of thing.
To a first approximation, implementing - lets call it machine automation and optimization - into how we do what we call F3EAD (Find, fix, finish, exploit, analyze, disseminate) will effectively do what we do now, only faster and more precisely.
Beyond that, it's pretty open as to what kind of novel planning capabilities will emerge from implementing DRL and other combined optimization strategies within kinetic and non-kinetic conflicts.
Note also, that the state space for force on force conflict, even small unit tactics are orders of magnitude more complex than something like GO because they are neither fully-observable nor deterministic (and can't easily be approximated as such).